Click here to load reader
Upload
phamkhanh
View
212
Download
0
Embed Size (px)
Citation preview
The employee-customersatisfaction chain in the ECSI
modelManuel JoseAcirc Vilares and Pedro SimoAumles CoelhoISEGI plusmn New University of Lisbon Lisbon Portugal
Keywords Customer satisfaction Career satisfaction Marketing models
Abstract Among the large number of currently available approaches for studying customersatisfaction a very promising one is that adopted in the European customer satisfaction index(ECSI) model Yet in spite of its various contributions to customer satisfaction research thisapproach exhibits certain limitations of which we will emphasise one contrary to compellingevidence the model does not consider the service climate or more speciregcally the cause and effectrelationship between employee behaviour and customer satisfaction The main goal of the presentpaper is to contribute to counteracting such a limitation A reformulation of the ECSI model issuggested integrating it into key components of employee satisfaction models (employeesatisfaction loyalty and commitment) as they are perceived by customers Both the ECSI modeland the ECSI revised model are estimated with data from a survey carried out among supermarketcustomers The results show that some variables such as perceived quality customer satisfactionand loyalty are better explained by the ECSI revised model Also statistically signiregcantinteractions between the new variables (with the exception of employee loyalty) and some of theECSI model variables (perceived product quality and perceived service quality) were discovered
IntroductionIn the 1990s customer satisfaction had a signiregcant impact on managementthinking In fact the realisation that understanding meeting and anticipatingcustomer needs was probably the most important source of sustained andcompetitive advantage for a company had a decisive effect on the setting ofcorporate priorities and practices
Among the large number of currently available approaches for studyingcustomer satisfaction a very promising one appears in the Swedish barometerin 1989 (Fornell 1992) It was followed in 1994 by the start-up of the Americancustomer satisfaction index (Fornell et al 1996) and more recently with thepreparation of the European customer satisfaction index (ECSI) (ECSITechnical Committee 1998) This approach computes a customer satisfactionindex using an econometric model that in terms of a causal relationship ties aset of latent variables (like customer expectations and customer perceptions ofquality and value) to a customer satisfaction index The model then ties thisindex to customer loyalty and other performance indicators
The Emerald Research Register for this journal is available at The current issue and full text archive of this journal is available at
httpwww emeraldinsight comresearchregister http wwwemeraldinsigh tcom0309-056 6htm
The authors would like to thank the anonymous referees for their helpful comments
Employee-customer
satisfaction
1703
Received June 2000Revised March 2001June 2001 September
2001
European Journal of MarketingVol 37 No 1112 2003
pp 1703-1722
q MCB UP Limited0309-0566
DOI 10110803090560310495429
However this approach exhibits certain limitations of which we willemphasise one the model does not consider the service climate or morespeciregcally the cause and effect relationship between employee behaviour andcustomer satisfaction As will be emphasized the fact that the ECSI model doesnot consider employee satisfaction and behaviour represents a limitation Themain goal of the present paper is to contribute to counteracting such alimitation
The paper is organised as follows The next section surveys previous workregarding the employee-customer link In the following section a reformulationof the ECSI model is proposed integrating some of the components of theemployee satisfaction models as they are perceived by customers The nextsection deals with methodological questions on measuring customer andemployee satisfaction and with the modelling approaches adopted to overcomesuch difregculties The estimation method of this revised ECSI model ispresented in the following section while in the penultimate section there is ananalysis of the results of the estimation of the model with data from a surveycarried out among supermarket customers The main conclusions are presentedin the regnal section
Employee-customer linksSeveral empirical studies (see a survey in Schmit and Allscheid (1995)) showthat it is impossible to maintain a satisreged and loyal customer base withoutsatisreged and loyal employees The studies show a signiregcant impact oncustomer satisfaction following an improvement in employee attitudes
The linkage between customer and employee variables has also beendepicted by Hesket et al (1997) within a framework termed the service proregtchain or employee-customer proregt chain According to this framework theelements of the chain are interdependent and complex and the size and strengthof their inmacruence varies by industry market segment and even organizationalfunction Nevertheless the basic links can be described as follows
employee variables like employee satisfaction commitment and loyaltyinmacruence customer perception of the value of the product and servicewhich in turn inmacruences customer satisfaction
customer satisfaction inmacruences customer loyalty and
corporate regnancial results are directly inmacruenced by customer loyalty
Following this approach[1] Rucci et al (1998) analysed the employee-customerproregt chain at the company Sears Roebuck and Co In their model employeebehaviour is explained by three variables attitude about the job attitude aboutthe company and employee retention They have estimated that a regve-pointimprovement in employee attitudes drives a 13 point rise in customersatisfaction which in turn will drive a 05 per cent improvement in revenuegrowth As is explicitly emphasised (Rucci et al 1998 p 91) ordfThese numbers
EJM371112
1704
are as rigorous as any other numbers we work at Sears Every year ouraccounting regrm audits them as closely as it audits our regnancialsordm
More recently Brooks (2000) overviews the research on the relationshipsbetween regnancial success and customer and employee variables According tothis research between 40 and 80 per cent of customer satisfaction and loyalty isdetermined by the customer-employee relationship depending on the industryand market segment that is being considered
On the other hand as Crosby et al (1994 p 21) point out ordfIf employees aretruly motivated by a desire to do quality work that meets customer needs thenachievement of that outcome (customer satisfaction) should contribute to theirown satisfaction as wellordm
So compelling evidence shows that there is a strong linkage betweenemployee and customer satisfactions
Employee-customer links in the ECSI modelThis section proposes an extension of the ECSI model[2] resulting from theinclusion of three new latent variables (perceived employee satisfactionperceived employee loyalty and perceived employee commitment)[3] thatrepresent employee satisfaction and behaviour as perceived by the customer[4]This choice is based on two kinds of assumptions on the one hand these arethe variables regarding employee satisfaction and behaviour that are likely toexplain customer satisfaction and on the other hand these are the variablesthat can be most easily perceived by the customers Nevertheless thesecustomer perceptions about employee satisfaction loyalty and commitment arenot necessarily correct However they can be good proxies for the ordftrueordmemployee variables particularly in companies where there are close andregular contacts between customers and employees In the regnal section we willgo further on this point
The extended model is shown in Figure 1[5] The rectangles represent latentvariables which are associated with a number of manifest variables not shownin the reggure The lines represent casual relationships The model includes nineinterrelated latent variables with two exogenous variables image andperceived employee satisfaction
Perceived employee satisfactionThis variable represents the way customers perceive employeesrsquo satisfactionThis satisfaction represents feelings of the employee about the job deregned asthe overall evaluation of working for the company Supposing that customerperceptions are not excessively wrong then one can admit that theseperceptions have a positive impact on customer satisfaction and loyalty Thisimpact is supposed to act indirectly through the impact on perceived quality(particularly on the service component)
Employee-customer
satisfaction
1705
Figure 1Extended customersatisfaction model
EJM371112
1706
We have also considered impacts from perceived employee satisfaction onperceived employee loyalty and commitment in conformity with employeesatisfaction models
Perceived employee loyaltyThis variable represents the way customers perceive employee loyalty Thisloyalty means the employeersquos intention to remain with the company andwillingness to recommend the company as a good place to work
We have admitted a possible direct impact from perceived employee loyaltyon customer loyalty This impact may exist mainly in companies where thereis personal contact between customers and employees
In fact it is well known that employee loyalty may have a positive effect oncustomer loyalty which in turn is a key determinant of proregtability incompanies (Reichheld and Sasser 1990) As Syrett (1997 p 49) points out ordfStaffloyalty is not a new business concern What makes the current debate differentis that in the age of stakeholder management and total quality managementcompanies have started to make links between the loyalty of their staff and thecorresponding loyalty of their customers and investorsordm
In conformity with employee satisfaction models an impact from perceivedemployee loyalty on perceived employee commitment is also considered
Perceived employee commitmentThis variable represents the way customers perceive employee commitment Itis well known that business success requires more than just satisreged and loyalemployees Instead it demands the kind of employees who are willing to serveas advocates for the organisation ie committed employees So employeecommitment represents employee dedication to help the company to achieve itsgoals It includes manifests like dedication to doing work of high qualitycommitment to resolving customersrsquo problems the investment of adequate timeand effort in the work and the will to recommend the companyrsquos products andservices
Though related employment commitment and loyalty are different anddistinct constructs with different implications following a change in acompanyrsquos focus[6] Also customersrsquo perceptions about employee loyalty andcommitment may be in some companies quite different The effects producedby these two constructs are expected to be of a different nature In fact whileperceived employee commitment will have its main inmacruence on perceivedquality (product and service) in the case of perceived employee loyalty thiseffect is not expected but rather a possible direct impact on customer loyaltyFor this reason the proposed model includes perceived employee loyalty andperceived employee commitment as two different latent variables andconsiders direct impacts from perceived employee commitment on perceivedproduct quality and perceived service quality
Employee-customer
satisfaction
1707
Methodological issues and modelling approachesThe measurement of customer and employee satisfaction and the modelling oftheir determinants give rise to a signiregcant number of methodological issues
Unlike income and proregts satisfaction is viewed as a latent construct thatis not observed directly and can only be estimated through indicatorsMoreover there is no single concept of satisfaction In fact according toAnderson et al (1994) at least two different conceptualisations can bedistinguished transaction speciregc and cumulative From a transactionspeciregc perspective customer satisfaction is viewed as a post-choiceevaluative judgement of a speciregc purchase occasion (Oliver and Swan1989) In contrast cumulative customer satisfaction is an overallevaluation based on the total purchase and consumption of a product orservice over time (Fornell 1992) Nevertheless much work has been donein trying to adequately measure customer satisfaction[7] In particularthe indicators adopted in the ECSI model to measure customersatisfaction were those recommended by Ryan et al (1995) overallsatisfaction fulreglment of expectations and distance to the customerrsquosideal company
These indicators may correspond to questionnaire items (manifest ormeasurement variables) or they may themselves be latent constructs likeexpectations and perceived quality In both cases the indicators dependon the area of application and it is not an easy task to set up a generalframework that encompasses the companies in all sectors For instancecustomer repurchasing has a quite different meaning when it is used witha company in a competitive market than when it is used with a companyor public administration producing a public good or service in amonopolistic situation
The questionnaire items are difregcult to deregne and they are usuallymeasured in terms of rating scales that try to capture the strength of anattitude (in the ECSI approach a ten-point numerical rating scale is used)On the other hand the questionnaire-item responses usually have skewedand non-normal distributions
The relationships between questionnaire items on the one hand andsatisfaction and other latent constructs (such as attitudes expectationsand perceived performances) on the other are not straightforward andthey have to be specireged with an error term
The questionnaires used to estimate employee satisfaction gatherinformation from surveys of the employeesmanagers of the companiesanalysed while the customer satisfaction models use customer surveysSo when analysing the employee-customer satisfaction chain one shouldpay attention to the fact that the entities observed in the two cases aredifferent
EJM371112
1708
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
However this approach exhibits certain limitations of which we willemphasise one the model does not consider the service climate or morespeciregcally the cause and effect relationship between employee behaviour andcustomer satisfaction As will be emphasized the fact that the ECSI model doesnot consider employee satisfaction and behaviour represents a limitation Themain goal of the present paper is to contribute to counteracting such alimitation
The paper is organised as follows The next section surveys previous workregarding the employee-customer link In the following section a reformulationof the ECSI model is proposed integrating some of the components of theemployee satisfaction models as they are perceived by customers The nextsection deals with methodological questions on measuring customer andemployee satisfaction and with the modelling approaches adopted to overcomesuch difregculties The estimation method of this revised ECSI model ispresented in the following section while in the penultimate section there is ananalysis of the results of the estimation of the model with data from a surveycarried out among supermarket customers The main conclusions are presentedin the regnal section
Employee-customer linksSeveral empirical studies (see a survey in Schmit and Allscheid (1995)) showthat it is impossible to maintain a satisreged and loyal customer base withoutsatisreged and loyal employees The studies show a signiregcant impact oncustomer satisfaction following an improvement in employee attitudes
The linkage between customer and employee variables has also beendepicted by Hesket et al (1997) within a framework termed the service proregtchain or employee-customer proregt chain According to this framework theelements of the chain are interdependent and complex and the size and strengthof their inmacruence varies by industry market segment and even organizationalfunction Nevertheless the basic links can be described as follows
employee variables like employee satisfaction commitment and loyaltyinmacruence customer perception of the value of the product and servicewhich in turn inmacruences customer satisfaction
customer satisfaction inmacruences customer loyalty and
corporate regnancial results are directly inmacruenced by customer loyalty
Following this approach[1] Rucci et al (1998) analysed the employee-customerproregt chain at the company Sears Roebuck and Co In their model employeebehaviour is explained by three variables attitude about the job attitude aboutthe company and employee retention They have estimated that a regve-pointimprovement in employee attitudes drives a 13 point rise in customersatisfaction which in turn will drive a 05 per cent improvement in revenuegrowth As is explicitly emphasised (Rucci et al 1998 p 91) ordfThese numbers
EJM371112
1704
are as rigorous as any other numbers we work at Sears Every year ouraccounting regrm audits them as closely as it audits our regnancialsordm
More recently Brooks (2000) overviews the research on the relationshipsbetween regnancial success and customer and employee variables According tothis research between 40 and 80 per cent of customer satisfaction and loyalty isdetermined by the customer-employee relationship depending on the industryand market segment that is being considered
On the other hand as Crosby et al (1994 p 21) point out ordfIf employees aretruly motivated by a desire to do quality work that meets customer needs thenachievement of that outcome (customer satisfaction) should contribute to theirown satisfaction as wellordm
So compelling evidence shows that there is a strong linkage betweenemployee and customer satisfactions
Employee-customer links in the ECSI modelThis section proposes an extension of the ECSI model[2] resulting from theinclusion of three new latent variables (perceived employee satisfactionperceived employee loyalty and perceived employee commitment)[3] thatrepresent employee satisfaction and behaviour as perceived by the customer[4]This choice is based on two kinds of assumptions on the one hand these arethe variables regarding employee satisfaction and behaviour that are likely toexplain customer satisfaction and on the other hand these are the variablesthat can be most easily perceived by the customers Nevertheless thesecustomer perceptions about employee satisfaction loyalty and commitment arenot necessarily correct However they can be good proxies for the ordftrueordmemployee variables particularly in companies where there are close andregular contacts between customers and employees In the regnal section we willgo further on this point
The extended model is shown in Figure 1[5] The rectangles represent latentvariables which are associated with a number of manifest variables not shownin the reggure The lines represent casual relationships The model includes nineinterrelated latent variables with two exogenous variables image andperceived employee satisfaction
Perceived employee satisfactionThis variable represents the way customers perceive employeesrsquo satisfactionThis satisfaction represents feelings of the employee about the job deregned asthe overall evaluation of working for the company Supposing that customerperceptions are not excessively wrong then one can admit that theseperceptions have a positive impact on customer satisfaction and loyalty Thisimpact is supposed to act indirectly through the impact on perceived quality(particularly on the service component)
Employee-customer
satisfaction
1705
Figure 1Extended customersatisfaction model
EJM371112
1706
We have also considered impacts from perceived employee satisfaction onperceived employee loyalty and commitment in conformity with employeesatisfaction models
Perceived employee loyaltyThis variable represents the way customers perceive employee loyalty Thisloyalty means the employeersquos intention to remain with the company andwillingness to recommend the company as a good place to work
We have admitted a possible direct impact from perceived employee loyaltyon customer loyalty This impact may exist mainly in companies where thereis personal contact between customers and employees
In fact it is well known that employee loyalty may have a positive effect oncustomer loyalty which in turn is a key determinant of proregtability incompanies (Reichheld and Sasser 1990) As Syrett (1997 p 49) points out ordfStaffloyalty is not a new business concern What makes the current debate differentis that in the age of stakeholder management and total quality managementcompanies have started to make links between the loyalty of their staff and thecorresponding loyalty of their customers and investorsordm
In conformity with employee satisfaction models an impact from perceivedemployee loyalty on perceived employee commitment is also considered
Perceived employee commitmentThis variable represents the way customers perceive employee commitment Itis well known that business success requires more than just satisreged and loyalemployees Instead it demands the kind of employees who are willing to serveas advocates for the organisation ie committed employees So employeecommitment represents employee dedication to help the company to achieve itsgoals It includes manifests like dedication to doing work of high qualitycommitment to resolving customersrsquo problems the investment of adequate timeand effort in the work and the will to recommend the companyrsquos products andservices
Though related employment commitment and loyalty are different anddistinct constructs with different implications following a change in acompanyrsquos focus[6] Also customersrsquo perceptions about employee loyalty andcommitment may be in some companies quite different The effects producedby these two constructs are expected to be of a different nature In fact whileperceived employee commitment will have its main inmacruence on perceivedquality (product and service) in the case of perceived employee loyalty thiseffect is not expected but rather a possible direct impact on customer loyaltyFor this reason the proposed model includes perceived employee loyalty andperceived employee commitment as two different latent variables andconsiders direct impacts from perceived employee commitment on perceivedproduct quality and perceived service quality
Employee-customer
satisfaction
1707
Methodological issues and modelling approachesThe measurement of customer and employee satisfaction and the modelling oftheir determinants give rise to a signiregcant number of methodological issues
Unlike income and proregts satisfaction is viewed as a latent construct thatis not observed directly and can only be estimated through indicatorsMoreover there is no single concept of satisfaction In fact according toAnderson et al (1994) at least two different conceptualisations can bedistinguished transaction speciregc and cumulative From a transactionspeciregc perspective customer satisfaction is viewed as a post-choiceevaluative judgement of a speciregc purchase occasion (Oliver and Swan1989) In contrast cumulative customer satisfaction is an overallevaluation based on the total purchase and consumption of a product orservice over time (Fornell 1992) Nevertheless much work has been donein trying to adequately measure customer satisfaction[7] In particularthe indicators adopted in the ECSI model to measure customersatisfaction were those recommended by Ryan et al (1995) overallsatisfaction fulreglment of expectations and distance to the customerrsquosideal company
These indicators may correspond to questionnaire items (manifest ormeasurement variables) or they may themselves be latent constructs likeexpectations and perceived quality In both cases the indicators dependon the area of application and it is not an easy task to set up a generalframework that encompasses the companies in all sectors For instancecustomer repurchasing has a quite different meaning when it is used witha company in a competitive market than when it is used with a companyor public administration producing a public good or service in amonopolistic situation
The questionnaire items are difregcult to deregne and they are usuallymeasured in terms of rating scales that try to capture the strength of anattitude (in the ECSI approach a ten-point numerical rating scale is used)On the other hand the questionnaire-item responses usually have skewedand non-normal distributions
The relationships between questionnaire items on the one hand andsatisfaction and other latent constructs (such as attitudes expectationsand perceived performances) on the other are not straightforward andthey have to be specireged with an error term
The questionnaires used to estimate employee satisfaction gatherinformation from surveys of the employeesmanagers of the companiesanalysed while the customer satisfaction models use customer surveysSo when analysing the employee-customer satisfaction chain one shouldpay attention to the fact that the entities observed in the two cases aredifferent
EJM371112
1708
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
are as rigorous as any other numbers we work at Sears Every year ouraccounting regrm audits them as closely as it audits our regnancialsordm
More recently Brooks (2000) overviews the research on the relationshipsbetween regnancial success and customer and employee variables According tothis research between 40 and 80 per cent of customer satisfaction and loyalty isdetermined by the customer-employee relationship depending on the industryand market segment that is being considered
On the other hand as Crosby et al (1994 p 21) point out ordfIf employees aretruly motivated by a desire to do quality work that meets customer needs thenachievement of that outcome (customer satisfaction) should contribute to theirown satisfaction as wellordm
So compelling evidence shows that there is a strong linkage betweenemployee and customer satisfactions
Employee-customer links in the ECSI modelThis section proposes an extension of the ECSI model[2] resulting from theinclusion of three new latent variables (perceived employee satisfactionperceived employee loyalty and perceived employee commitment)[3] thatrepresent employee satisfaction and behaviour as perceived by the customer[4]This choice is based on two kinds of assumptions on the one hand these arethe variables regarding employee satisfaction and behaviour that are likely toexplain customer satisfaction and on the other hand these are the variablesthat can be most easily perceived by the customers Nevertheless thesecustomer perceptions about employee satisfaction loyalty and commitment arenot necessarily correct However they can be good proxies for the ordftrueordmemployee variables particularly in companies where there are close andregular contacts between customers and employees In the regnal section we willgo further on this point
The extended model is shown in Figure 1[5] The rectangles represent latentvariables which are associated with a number of manifest variables not shownin the reggure The lines represent casual relationships The model includes nineinterrelated latent variables with two exogenous variables image andperceived employee satisfaction
Perceived employee satisfactionThis variable represents the way customers perceive employeesrsquo satisfactionThis satisfaction represents feelings of the employee about the job deregned asthe overall evaluation of working for the company Supposing that customerperceptions are not excessively wrong then one can admit that theseperceptions have a positive impact on customer satisfaction and loyalty Thisimpact is supposed to act indirectly through the impact on perceived quality(particularly on the service component)
Employee-customer
satisfaction
1705
Figure 1Extended customersatisfaction model
EJM371112
1706
We have also considered impacts from perceived employee satisfaction onperceived employee loyalty and commitment in conformity with employeesatisfaction models
Perceived employee loyaltyThis variable represents the way customers perceive employee loyalty Thisloyalty means the employeersquos intention to remain with the company andwillingness to recommend the company as a good place to work
We have admitted a possible direct impact from perceived employee loyaltyon customer loyalty This impact may exist mainly in companies where thereis personal contact between customers and employees
In fact it is well known that employee loyalty may have a positive effect oncustomer loyalty which in turn is a key determinant of proregtability incompanies (Reichheld and Sasser 1990) As Syrett (1997 p 49) points out ordfStaffloyalty is not a new business concern What makes the current debate differentis that in the age of stakeholder management and total quality managementcompanies have started to make links between the loyalty of their staff and thecorresponding loyalty of their customers and investorsordm
In conformity with employee satisfaction models an impact from perceivedemployee loyalty on perceived employee commitment is also considered
Perceived employee commitmentThis variable represents the way customers perceive employee commitment Itis well known that business success requires more than just satisreged and loyalemployees Instead it demands the kind of employees who are willing to serveas advocates for the organisation ie committed employees So employeecommitment represents employee dedication to help the company to achieve itsgoals It includes manifests like dedication to doing work of high qualitycommitment to resolving customersrsquo problems the investment of adequate timeand effort in the work and the will to recommend the companyrsquos products andservices
Though related employment commitment and loyalty are different anddistinct constructs with different implications following a change in acompanyrsquos focus[6] Also customersrsquo perceptions about employee loyalty andcommitment may be in some companies quite different The effects producedby these two constructs are expected to be of a different nature In fact whileperceived employee commitment will have its main inmacruence on perceivedquality (product and service) in the case of perceived employee loyalty thiseffect is not expected but rather a possible direct impact on customer loyaltyFor this reason the proposed model includes perceived employee loyalty andperceived employee commitment as two different latent variables andconsiders direct impacts from perceived employee commitment on perceivedproduct quality and perceived service quality
Employee-customer
satisfaction
1707
Methodological issues and modelling approachesThe measurement of customer and employee satisfaction and the modelling oftheir determinants give rise to a signiregcant number of methodological issues
Unlike income and proregts satisfaction is viewed as a latent construct thatis not observed directly and can only be estimated through indicatorsMoreover there is no single concept of satisfaction In fact according toAnderson et al (1994) at least two different conceptualisations can bedistinguished transaction speciregc and cumulative From a transactionspeciregc perspective customer satisfaction is viewed as a post-choiceevaluative judgement of a speciregc purchase occasion (Oliver and Swan1989) In contrast cumulative customer satisfaction is an overallevaluation based on the total purchase and consumption of a product orservice over time (Fornell 1992) Nevertheless much work has been donein trying to adequately measure customer satisfaction[7] In particularthe indicators adopted in the ECSI model to measure customersatisfaction were those recommended by Ryan et al (1995) overallsatisfaction fulreglment of expectations and distance to the customerrsquosideal company
These indicators may correspond to questionnaire items (manifest ormeasurement variables) or they may themselves be latent constructs likeexpectations and perceived quality In both cases the indicators dependon the area of application and it is not an easy task to set up a generalframework that encompasses the companies in all sectors For instancecustomer repurchasing has a quite different meaning when it is used witha company in a competitive market than when it is used with a companyor public administration producing a public good or service in amonopolistic situation
The questionnaire items are difregcult to deregne and they are usuallymeasured in terms of rating scales that try to capture the strength of anattitude (in the ECSI approach a ten-point numerical rating scale is used)On the other hand the questionnaire-item responses usually have skewedand non-normal distributions
The relationships between questionnaire items on the one hand andsatisfaction and other latent constructs (such as attitudes expectationsand perceived performances) on the other are not straightforward andthey have to be specireged with an error term
The questionnaires used to estimate employee satisfaction gatherinformation from surveys of the employeesmanagers of the companiesanalysed while the customer satisfaction models use customer surveysSo when analysing the employee-customer satisfaction chain one shouldpay attention to the fact that the entities observed in the two cases aredifferent
EJM371112
1708
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Figure 1Extended customersatisfaction model
EJM371112
1706
We have also considered impacts from perceived employee satisfaction onperceived employee loyalty and commitment in conformity with employeesatisfaction models
Perceived employee loyaltyThis variable represents the way customers perceive employee loyalty Thisloyalty means the employeersquos intention to remain with the company andwillingness to recommend the company as a good place to work
We have admitted a possible direct impact from perceived employee loyaltyon customer loyalty This impact may exist mainly in companies where thereis personal contact between customers and employees
In fact it is well known that employee loyalty may have a positive effect oncustomer loyalty which in turn is a key determinant of proregtability incompanies (Reichheld and Sasser 1990) As Syrett (1997 p 49) points out ordfStaffloyalty is not a new business concern What makes the current debate differentis that in the age of stakeholder management and total quality managementcompanies have started to make links between the loyalty of their staff and thecorresponding loyalty of their customers and investorsordm
In conformity with employee satisfaction models an impact from perceivedemployee loyalty on perceived employee commitment is also considered
Perceived employee commitmentThis variable represents the way customers perceive employee commitment Itis well known that business success requires more than just satisreged and loyalemployees Instead it demands the kind of employees who are willing to serveas advocates for the organisation ie committed employees So employeecommitment represents employee dedication to help the company to achieve itsgoals It includes manifests like dedication to doing work of high qualitycommitment to resolving customersrsquo problems the investment of adequate timeand effort in the work and the will to recommend the companyrsquos products andservices
Though related employment commitment and loyalty are different anddistinct constructs with different implications following a change in acompanyrsquos focus[6] Also customersrsquo perceptions about employee loyalty andcommitment may be in some companies quite different The effects producedby these two constructs are expected to be of a different nature In fact whileperceived employee commitment will have its main inmacruence on perceivedquality (product and service) in the case of perceived employee loyalty thiseffect is not expected but rather a possible direct impact on customer loyaltyFor this reason the proposed model includes perceived employee loyalty andperceived employee commitment as two different latent variables andconsiders direct impacts from perceived employee commitment on perceivedproduct quality and perceived service quality
Employee-customer
satisfaction
1707
Methodological issues and modelling approachesThe measurement of customer and employee satisfaction and the modelling oftheir determinants give rise to a signiregcant number of methodological issues
Unlike income and proregts satisfaction is viewed as a latent construct thatis not observed directly and can only be estimated through indicatorsMoreover there is no single concept of satisfaction In fact according toAnderson et al (1994) at least two different conceptualisations can bedistinguished transaction speciregc and cumulative From a transactionspeciregc perspective customer satisfaction is viewed as a post-choiceevaluative judgement of a speciregc purchase occasion (Oliver and Swan1989) In contrast cumulative customer satisfaction is an overallevaluation based on the total purchase and consumption of a product orservice over time (Fornell 1992) Nevertheless much work has been donein trying to adequately measure customer satisfaction[7] In particularthe indicators adopted in the ECSI model to measure customersatisfaction were those recommended by Ryan et al (1995) overallsatisfaction fulreglment of expectations and distance to the customerrsquosideal company
These indicators may correspond to questionnaire items (manifest ormeasurement variables) or they may themselves be latent constructs likeexpectations and perceived quality In both cases the indicators dependon the area of application and it is not an easy task to set up a generalframework that encompasses the companies in all sectors For instancecustomer repurchasing has a quite different meaning when it is used witha company in a competitive market than when it is used with a companyor public administration producing a public good or service in amonopolistic situation
The questionnaire items are difregcult to deregne and they are usuallymeasured in terms of rating scales that try to capture the strength of anattitude (in the ECSI approach a ten-point numerical rating scale is used)On the other hand the questionnaire-item responses usually have skewedand non-normal distributions
The relationships between questionnaire items on the one hand andsatisfaction and other latent constructs (such as attitudes expectationsand perceived performances) on the other are not straightforward andthey have to be specireged with an error term
The questionnaires used to estimate employee satisfaction gatherinformation from surveys of the employeesmanagers of the companiesanalysed while the customer satisfaction models use customer surveysSo when analysing the employee-customer satisfaction chain one shouldpay attention to the fact that the entities observed in the two cases aredifferent
EJM371112
1708
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
We have also considered impacts from perceived employee satisfaction onperceived employee loyalty and commitment in conformity with employeesatisfaction models
Perceived employee loyaltyThis variable represents the way customers perceive employee loyalty Thisloyalty means the employeersquos intention to remain with the company andwillingness to recommend the company as a good place to work
We have admitted a possible direct impact from perceived employee loyaltyon customer loyalty This impact may exist mainly in companies where thereis personal contact between customers and employees
In fact it is well known that employee loyalty may have a positive effect oncustomer loyalty which in turn is a key determinant of proregtability incompanies (Reichheld and Sasser 1990) As Syrett (1997 p 49) points out ordfStaffloyalty is not a new business concern What makes the current debate differentis that in the age of stakeholder management and total quality managementcompanies have started to make links between the loyalty of their staff and thecorresponding loyalty of their customers and investorsordm
In conformity with employee satisfaction models an impact from perceivedemployee loyalty on perceived employee commitment is also considered
Perceived employee commitmentThis variable represents the way customers perceive employee commitment Itis well known that business success requires more than just satisreged and loyalemployees Instead it demands the kind of employees who are willing to serveas advocates for the organisation ie committed employees So employeecommitment represents employee dedication to help the company to achieve itsgoals It includes manifests like dedication to doing work of high qualitycommitment to resolving customersrsquo problems the investment of adequate timeand effort in the work and the will to recommend the companyrsquos products andservices
Though related employment commitment and loyalty are different anddistinct constructs with different implications following a change in acompanyrsquos focus[6] Also customersrsquo perceptions about employee loyalty andcommitment may be in some companies quite different The effects producedby these two constructs are expected to be of a different nature In fact whileperceived employee commitment will have its main inmacruence on perceivedquality (product and service) in the case of perceived employee loyalty thiseffect is not expected but rather a possible direct impact on customer loyaltyFor this reason the proposed model includes perceived employee loyalty andperceived employee commitment as two different latent variables andconsiders direct impacts from perceived employee commitment on perceivedproduct quality and perceived service quality
Employee-customer
satisfaction
1707
Methodological issues and modelling approachesThe measurement of customer and employee satisfaction and the modelling oftheir determinants give rise to a signiregcant number of methodological issues
Unlike income and proregts satisfaction is viewed as a latent construct thatis not observed directly and can only be estimated through indicatorsMoreover there is no single concept of satisfaction In fact according toAnderson et al (1994) at least two different conceptualisations can bedistinguished transaction speciregc and cumulative From a transactionspeciregc perspective customer satisfaction is viewed as a post-choiceevaluative judgement of a speciregc purchase occasion (Oliver and Swan1989) In contrast cumulative customer satisfaction is an overallevaluation based on the total purchase and consumption of a product orservice over time (Fornell 1992) Nevertheless much work has been donein trying to adequately measure customer satisfaction[7] In particularthe indicators adopted in the ECSI model to measure customersatisfaction were those recommended by Ryan et al (1995) overallsatisfaction fulreglment of expectations and distance to the customerrsquosideal company
These indicators may correspond to questionnaire items (manifest ormeasurement variables) or they may themselves be latent constructs likeexpectations and perceived quality In both cases the indicators dependon the area of application and it is not an easy task to set up a generalframework that encompasses the companies in all sectors For instancecustomer repurchasing has a quite different meaning when it is used witha company in a competitive market than when it is used with a companyor public administration producing a public good or service in amonopolistic situation
The questionnaire items are difregcult to deregne and they are usuallymeasured in terms of rating scales that try to capture the strength of anattitude (in the ECSI approach a ten-point numerical rating scale is used)On the other hand the questionnaire-item responses usually have skewedand non-normal distributions
The relationships between questionnaire items on the one hand andsatisfaction and other latent constructs (such as attitudes expectationsand perceived performances) on the other are not straightforward andthey have to be specireged with an error term
The questionnaires used to estimate employee satisfaction gatherinformation from surveys of the employeesmanagers of the companiesanalysed while the customer satisfaction models use customer surveysSo when analysing the employee-customer satisfaction chain one shouldpay attention to the fact that the entities observed in the two cases aredifferent
EJM371112
1708
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Methodological issues and modelling approachesThe measurement of customer and employee satisfaction and the modelling oftheir determinants give rise to a signiregcant number of methodological issues
Unlike income and proregts satisfaction is viewed as a latent construct thatis not observed directly and can only be estimated through indicatorsMoreover there is no single concept of satisfaction In fact according toAnderson et al (1994) at least two different conceptualisations can bedistinguished transaction speciregc and cumulative From a transactionspeciregc perspective customer satisfaction is viewed as a post-choiceevaluative judgement of a speciregc purchase occasion (Oliver and Swan1989) In contrast cumulative customer satisfaction is an overallevaluation based on the total purchase and consumption of a product orservice over time (Fornell 1992) Nevertheless much work has been donein trying to adequately measure customer satisfaction[7] In particularthe indicators adopted in the ECSI model to measure customersatisfaction were those recommended by Ryan et al (1995) overallsatisfaction fulreglment of expectations and distance to the customerrsquosideal company
These indicators may correspond to questionnaire items (manifest ormeasurement variables) or they may themselves be latent constructs likeexpectations and perceived quality In both cases the indicators dependon the area of application and it is not an easy task to set up a generalframework that encompasses the companies in all sectors For instancecustomer repurchasing has a quite different meaning when it is used witha company in a competitive market than when it is used with a companyor public administration producing a public good or service in amonopolistic situation
The questionnaire items are difregcult to deregne and they are usuallymeasured in terms of rating scales that try to capture the strength of anattitude (in the ECSI approach a ten-point numerical rating scale is used)On the other hand the questionnaire-item responses usually have skewedand non-normal distributions
The relationships between questionnaire items on the one hand andsatisfaction and other latent constructs (such as attitudes expectationsand perceived performances) on the other are not straightforward andthey have to be specireged with an error term
The questionnaires used to estimate employee satisfaction gatherinformation from surveys of the employeesmanagers of the companiesanalysed while the customer satisfaction models use customer surveysSo when analysing the employee-customer satisfaction chain one shouldpay attention to the fact that the entities observed in the two cases aredifferent
EJM371112
1708
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
There is a time lag between a change in the climate in the organisation(particularly in employee satisfaction) and its effect on customersatisfaction and another before customers start to affect employees Suchtime lags are not observed
Data on employees obtained from customer surveys can only includecustomer perceptions about employeesrsquo attitudes and behaviours In ourmodel it is suggested to capture perceived employee satisfaction loyaltyand commitment by only asking customers
Resolving these methodological questions led naturally to different modellingapproaches to employee and customer satisfaction
An important classiregcation of these approaches distinguishes betweenstructural equation models and unstructured approaches
The second approach is probably the more commonly used It is widelyadopted in different studies (see various references in the work of Schmit andAllscheld (1995)) This type of unstructured approach is only descriptive andinvolves no proper structure or parameter estimates The correlation betweenthe different questionnaire items and dimensions is estimated through factor orcluster analysis In this approach there is no causal relationship or any otherkind between these different items and dimensions It is not possible to analyseinteraction and derive any cause and effect relationships and consequently itwill not be followed in the present research
The structural equation models presuppose a set of assumptions about thecustomer or employee decision process that are expressed through the numberand speciregcation of the equations and through the estimates for theparameters If a probabilistic component is integrated in the model it ispossible to give precision measurements of the parameter estimates and of thequality of the results obtained
These models are however complex and difregcult to estimate especiallywhen a simultaneous system approach is adopted Moreover given the skewand non-normality of the variables stated earlier the conventional maximumlikelihood estimators are not to be suggested
In an attempt to keep things simple at least initially some authors estimatesingle equation models Given the complex structure of the decision-makingprocess of both employees and customers a preliminary factor analysis iscarried out before the estimation of the single equation model This analysisallows us to select and investigate manifest variables (ie questionnaire items)in some detail before any consideration of latent constructs There are threedirect indicators of employee satisfaction (job satisfaction own morale andcolleaguesrsquo morale) that are considered as depending on a certain number ofquestionnaire items Given the large number of these items the authors carryout a preliminary factor analysis in order to reduce their number and to choosethose that are relatively unrelated
Employee-customer
satisfaction
1709
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
However this approach does not apply in the present study since the ECSImodel has several structural equations So in spite of the difregculties ofestimation our analysis is restricted to the structural simultaneous equationmodels The estimation method for the extended structural model is analysedin the next section
The estimation methodThe complete model which includes an inner structural model andmeasurement model is formally presented in the Appendix The structuralmodel is composed of nine latent variables as shown in Figure 1 Themeasurement model relates latent variables to the manifest variables
For the supermarket application in the following section the manifestvariables used with perceived employee satisfaction loyalty and commitmentare respectively
x21 plusmn perceived overall employee satisfaction
x22 plusmn distance to ideal place to work
y21 plusmn perceived employeesrsquo intentions to stay in the company
y22 plusmn perceived employeesrsquo intentions to plan a career in the company
y31 plusmn perceived employeesrsquo effort put into the job
y32 plusmn perceived employeesrsquo willingness to solve customersrsquo problems
y33 plusmn perceived quality of employeesrsquo work and
y34 plusmn perceived employeesrsquo willingness to recommend the products oftheir company
Some of the major difregculties in estimating this model are the presence of latent variables that are not observable and the fact of measuring categorical variables with an unknown non-normal
frequency distribution which is usually negatively skewed (Fornell1992)
Consequently a maximum likelihood approach (like LISREL) may not beappropriate for the problem Some authors have proposed partial least squares(PLS) which is already being used in the ESCI project (see a comparisonbetween maximum likelihood and PLS estimator properties in Dijkstra (1983))PLS was used to estimate the ECSI model and the revised ECSI model
An application to the case of supermarketsDataThe data used in the estimation come from a representative survey ofsupermarket customers In the following we will use data from twosupermarkets that we will refer to as Company A and Company B The regrstcomprises a number of big stores and the second small and medium stores The
EJM371112
1710
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
selection of the respondents follows the criteria deregned in ECSI TechnicalCommittee (1998)[8] The questionnaire used in the survey regards the overallexperience of the respondent with the supermarket and is also based on the oneadopted in the ECSI pilot project with the integration of the manifests for thethree new latent variables (see subsection ordfPerceived employee satisfactionordm) Sobesides questions regarding the seven constructs of the ECSI model (imageexpectations perceived product quality perceived service quality perceivedvalue customer satisfaction and customer loyalty) the questionnaire includes aset of questions regarding customer perceptions of employee satisfaction loyaltyand commitment The sample size is 298 for Company A and 249 for Company B
ResultsThe results presented in this section come from the estimation of two modelsthe ECSI model and the ECSI revised model given by equations (A3) and (A2)in the Appendix
Tables I-IV and Figures 2 and 3 show the main results Table I shows thevalues of the determination coefregcient (R 2) for the estimation of the equationsexplaining the latent variables Z4 (perceived product quality) Z5 (perceivedservice quality) Z6 (perceived value) Z7 (customer satisfaction) and Z8
(customer loyalty) Results show that for both companies all the latentvariables tend to be better explained by the ECSI revised model than by theECSI model This is particularly evident in both cases for perceived productquality and perceived service quality which show very important increases inR 2 Also the explanatory power of equations explaining customer satisfactionand customer loyalty show an improvement that is signiregcant for CompanyB For Company A we can still observe a small increase in customer loyalty
Company A Company BEquation ECSI ECSI revised ECSI ECSI revised
Z4 (Perceived product quality) 234 395 205 303Z5 (Perceived service quality) 304 583 343 449Z6 (Perceived value) 599 612 444 444Z7 (Customer satisfaction) 600 591 339 393Z8 (Customer loyalty) 514 535 401 450
Table IDetermination
coefregcients
Company A Company BVariable H 2 F 2 H 2 F 2
x2 (Perceived employee satisfaction) 050 000 080 000Z2 (Perceived employee loyalty) 075 023 072 027Z3 (Perceived employee commitment) 062 008 075 027
Table IICommunalities and
redundancy
Employee-customer
satisfaction
1711
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
69plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
620
28plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
053
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
035
008
029
020
037
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
051
009
033
020
038
000
plusmnplusmn
plusmnP
erce
ived
val
ue
030
006
021
025
048
022
045
plusmnplusmn
Sat
isfa
ctio
n0
220
040
160
390
290
280
240
28plusmn
Loy
alty
017
006
011
032
020
019
016
019
067
Table IIITotal effects(Company A)
EJM371112
1712
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Var
iable
Per
ceiv
edem
plo
yee
sati
sfac
tion
Per
ceiv
edem
plo
yee
loyal
ty
Per
ceiv
edem
plo
yee
com
mit
men
tIm
age
Expec
tati
ons
Per
ceiv
edpro
duct
qual
ity
Per
ceiv
edse
rvic
equal
ity
Per
ceiv
edval
ue
Sat
isfa
ctio
n
Per
ceiv
edem
plo
yee
sati
sfac
tion
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
loyal
ty0
64plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
emplo
yee
com
mit
men
t0
430
14plusmn
plusmnplusmn
plusmnplusmn
plusmnplusmn
Imag
e0
000
000
00plusmn
plusmnplusmn
plusmnplusmn
plusmnE
xpec
tati
ons
000
000
000
042
plusmnplusmn
plusmnplusmn
plusmnP
erce
ived
pro
duct
qual
ity
021
004
028
014
034
plusmnplusmn
plusmnplusmn
Per
ceiv
edse
rvic
equal
ity
025
004
031
019
044
000
plusmnplusmn
plusmnP
erce
ived
val
ue
014
002
017
018
042
002
054
plusmnplusmn
Sat
isfa
ctio
n0
090
020
120
410
200
190
210
10plusmn
Loy
alty
002
20
060
080
280
140
130
150
070
70
Table IVTotal effects(Company B)
Employee-customer
satisfaction
1713
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Figure 2Model parameterestimates and t values(Company A)
EJM371112
1714
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Figure 3Model parameter
estimates and t values(Company B)
Employee-customer
satisfaction
1715
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
explanatory power but the explanatory power of customer satisfactionremains almost unchanged (there is even a negligible decrease)
Table II shows the communalities (H 2) and redundancy coefregcients (F 2) forthe three employee variables included[9] These measures can be used asindicators for the validity and reliability of the measurement model It can be seenthat communalities for the three new variables are always signiregcantly higherthan 05 indicating that the variance captured by each latent variable is largerthan variance due to measurement error and thus demonstrating a high reliabilityof the construct The only exception is related to perceived employee satisfactionin Company A where the communality is equal to 05 This is due to the fact thatthe variable is almost completely determined by the indicator distance to idealplace to work Nevertheless considered globally results show a high reliabilityfor the indicators and constructs It can also be seen that redundancy coefregcientsare always signiregcantly smaller than communalities Values are never higherthan 027 indicating a low redundancy among the three proposed variablesConsequently they tend to conregrm the discriminant validity of the constructsAlso note that the squared correlations between the three variables are alwayssmaller than 032 for Company A and 04 for Company B The comparison ofthose values with communalities also satisreges the requirements for discriminantvalidity as proposed by Fornell and Larcker (1981)
Figures 2 and 3 show the estimates and t values for the model parameters forboth companies
Regarding the parameters that are common to the ECSI model and to theECSI revised model it can be observed that the interpretation of estimates isquite similar for both companies They show the following in both cases
Image has no direct impact on customer loyalty The impact is onlyindirect through customer satisfaction
Customer expectations have no direct impact on customer satisfaction Itis only indirect through perceived quality and perceived value
Perceived service quality although contributing to the explanation ofperceived value seems to have no direct impact on customer satisfactionNevertheless for Company B the introduction of employee variablesgenerates a statistically signiregcant impact from perceived service qualityon customer satisfaction This result shows that the consideration ofemployee variables may even contribute to a better estimation of therelationships between customer variables
Regarding the estimation of parameters speciregc to the ECSI revised modelresults are also highly consistent for both companies
Perceived employee satisfaction has important direct impact on perceivedservice quality for both companies the impact on perceived productquality is always weaker and even non-signiregcant for one company
EJM371112
1716
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Perceived employee commitment has important direct impact on perceivedproduct quality and perceived service quality for both companies also inthis case the impact on perceived service quality is stronger although thedifference between the two impacts is non-signiregcant
Globally it can be seen that employee variables signiregcantly contribute toexplain perceived quality
Effects of perceived employee commitment on perceived quality (product andservice) are always stronger than those originated by perceived employeesatisfaction it should be noted that perceived employee commitment seems tohave in both cases a very important explanatory capacity for perceivedquality almost of the same magnitude as expectations
The effect of perceived employee loyalty on customer loyalty isnon-signiregcant for both companies thus results show no empiricalevidence to sustain the assumption that there is a potential direct effectbetween these two variables for the companies studied
Tables III and IV show the total model effects (direct plus indirect effects) foreach company (origins of the effects in columns and destinations in rows)Results show the following
Perceived employee satisfaction and perceived employee commitmentconregrm very important explanatory capacity for perceived quality (bothproduct and service but mainly for the latter) for Company A the totalimpact originated by perceived employee satisfaction on perceivedservice quality is even stronger than that originated by expectations alsothese impacts are always stronger than that originated by image
The total impacts of perceived employee satisfaction and perceivedemployee commitment on customer satisfaction and loyalty althoughless strong are still signiregcant it should be noted that they tend to showmagnitudes very similar to those produced by perceived value
Perceived employee loyalty always shows very weak total effects both oncustomer satisfaction and customer loyalty the result is that for these twocompanies this variable does not bring signiregcant explanatory power forthe whole model
ConclusionsThe goal of this paper was to include the cause and effect relationship betweenemployee and customer satisfaction in the ECSI model
In this context we propose an ECSI revised model that is the ECSI modelwith customer perceptions on employee satisfaction loyalty and commitmentintegrated into it
The results of the estimation with data from surveys conducted amongsupermarket customers show that these latent variables have signiregcant
Employee-customer
satisfaction
1717
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
impacts on perceived quality (product and service) and signiregcantly contributeto explaining their variance Despite adding signiregcant explanatory power forperceived quality their contribution to explaining customer satisfaction andloyalty is somewhat lower This is not surprising since their effect on customersatisfaction and loyalty is mainly indirect (through perceived quality) andconsequently the increase in the explanatory power is only achieved by theircontribution to a better estimation of other customer variables Neverthelessthese new variables still show some signiregcant effects on customer satisfactionand loyalty In addition the inclusion of these employee variables showed thatit is possible to
(1) put in evidence the existence of an employee-customer chain
(2) quantify the effects of changes in employee attitudes and behaviours oncustomer satisfaction and loyalty (on the assumption that those changesare well perceived by customers) and
(3) improve the estimation of some ESCI model parameters and thus contributeto a better understanding of satisfaction and loyalty determinants
This research however is not complete since it is still necessary to analyse therelationship between employeesrsquo perceptions of customer satisfaction andloyalty on the one hand and employee satisfaction loyalty and commitmenton the other Such analysis would require the speciregcation and estimation of anemployee satisfaction model (with data from employee surveys)
In the context of two models (an employee model and a customer model) thekey question would then be to know if employees perceive customer satisfactionand loyalty properly and if customers for their part perceive employeesatisfaction loyalty and commitment properly It would also be interesting tounderstand how long it takes to form correct perceptions Thus the appropriateapproach would be to estimate both models simultaneously (the customer modeland the employee model) to compare the ordfrealordm values with the perceived valuesand to estimate the time lag involved in such relationships In a certain way itwould be following a similar approach to that of Schneider et al (1998) and itwould require cross-lagged panel analysis relating the ordfrealordm with the perceivedvalues of the variables Consequently it would require for a whole set ofperiods both data from customer surveys (for estimating the equations of thecustomer model) and similar data from employee surveys (for the employeemodel) Having no data available on employee surveys we have restricted theestimation in this paper to the customer satisfaction model ie thereformulated ECSI model using customer perceptions about employeevariables This obviously constitutes a major limitation of the study
Notes
1 Sergeant and Frenkel (2000) also focus on some aspects of the service-proregt chain exploringthe relationships between a number of variables Using a structural equation modelling
EJM371112
1718
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
methodology the authors found that in terms of direct effects on employeesrsquo capacity tosatisfy customers some variables (eg other department support) were more important thanothers (eg technology)
2 This is the model adopted in the 1999 ECSI wave
3 These variables are very close to the concepts of employee satisfaction loyalty andcommitment proposed by Crosby et al (1994)
4 Gremier and Gwinner (2000) also examine the customer-employee relationship from theperspective of the customer In fact they study the customer-employee rapport which isderegned as a customerrsquos perception of having an enjoyable interaction with a service provideremployee characterized by a personnel connection between the two interactants Theirresults show that both rapport components (enjoyable interaction and personnel connection)are signiregcantly related to customer satisfaction with the service and customer loyaltyintention as well as to the likelihood of positive customer communication about the regrm
5 Vilares and Coelho (1999) survey the employee satisfaction structural equation models andpropose an employee satisfaction model which adopts a similar framework to ECSI Some ofthe components of this model are integrated into the proposed reformulation of the ECSImodel
6 Crosby et al (1994) show that a change in a companyrsquos employee focus has a larger impact onloyalty than on commitment They have also pointed out that a change in a companyrsquosquality or customer focus has a larger impact on commitment than on loyalty
7 A comprehensive book on customer satisfaction measurement appeared in the publication ofVavra (1997)
8 The sampling design is based on the random selection of households using random digitdialling (RDD) In each household the selection of a resident is also made randomly The regrstset of questions in the questionnaire is used to qualify the potential respondent as a client ofthe sector (supermarkets) and of a particular supermarket chain All the other questions inthe questionnaire refer to the identireged supermarket The response rate was about 40 percent
9 Communality for a manifest variable may be interpreted as the proportion of its variancewhich is reproducedby the directly connected latent variable The redundancycoefregcient fora manifest variable is the proportion of its variance which is reproduced by the predictors ofits own latent variable Communalities and redundancy coefregcients for latent variables areaverages of the communalities and redundancy of their manifests
References
Anderson EW Fornel C and Lehmann DR (1994) ordfCustomer satisfaction market share andproregtabilityordm Journal of Marketing Vol 56 pp 53-66
Brooks R (2000) ordfWhy loyal employees and customers improve the bottom lineordm Journal ofQuality and Participation Vol 23 No 2 pp 40-4
Crosby LA Grisaffe DB and Marra TR (1994) ordfThe impact of quality and customersatisfaction on employee organisational commitmentordm Marketing and Research TodayFebruary pp 19-30
Dijkstra T (1983) ordfSome comments on maximum likelihood and partial least squares methodsordmJournal of Econometrics Vol 22 pp 67-90
ECSI Technical Committee (1998) European Customer Satisfaction Index Foundation andStructure for Harmonized National Pilot Projects ECSI
Fornell C (1992) ordfA national customer satisfaction barometer the Swedish experienceordm Journalof Marketing Vol 56 pp 6-21
Employee-customer
satisfaction
1719
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
Fornell C and Larcker DF (1981) ordfEvaluating structural equation models with unobservablevariables and measurement errorordm Journal of Marketing Vol 18 pp 39-50
Fornell C Johnson MD Anderson EW Cha J and Bryant BE (1996) ordfThe Americancustomer satisfaction index nature purpose and regndingsordm Journal of Marketing Vol 60pp 7-18
Gremier DG and Gwinner KP (2000) ordfCustomer-employee rapport in service relationshipsordmJournal of Service Research Vol 3 No 1 pp 82-104
Heskett JL Sasser WE and Schlesinger LA (1997) The Service Proregt Chain How LeadingCompanies Link Proregt and Growth to Loyalty Satisfaction and Value Free Press NewYork NY
Oliver RD and Swan JE (1989) ordfConsumer perceptions of interpersonal equity andsatisfaction in transactions a regeld survey approachordm Journal of Marketing Vol 53pp 21-35
Reichheld FF and Sasser WE (1990) ordfZero defections quality comes to serviceordm HarvardBusiness Review Vol 68 No 5 pp 105-11
Rucci AJ Kim SP and Quinn RT (1998) ordfThe employee-customer-proregt-chain at SearsordmHarvard Business Review Vol 76 No 1 pp 83-97
Ryan M Buzas T and Ramaswamy V (1995) ordfMaking CSM a power toolordm MarketingResearch Vol 7 No 3 pp 11-16
Schmit MS and Allscheid SP (1995) ordfEmployee attitudes and customer satisfaction makingtheoretical and empirical connectionsordm Personnel Psychology Vol 48 pp 521-36
Schneider B White SS and Paul MC (1998) ordfLinking service climate and customerperception of service quality test of a causal modelordm Journal of Applied Psychology Vol 8No 2 pp 150-63
Sergeant A and Frenkel S (2000) ordfWhen do customer contact employees satisfy customersordmJournal of Service Research Vol 3 No 1 pp 18-34
Syrett M (1997) ordfGoodbye to macho managementordm Director Vol 50 No 8 pp 49-56
Vavra TG (1997) Improving Your Measurement of Customer Satisfaction ASQ Quality PressMilwaukee WI
Vilares MJ and Coelho PS (1999) ordfThe employee-customer satisfaction chainordm paperpresented at the Conference on Customer Satisfaction Theory and Measurement Vienna
Further reading
Chang JS Labovitz G and Rosansky V (1993) Making Quality Work Harper Collins NewYork NY
Fosam EB Grimsley MFJ and Wisher SJ (1998) ordfExploring models for employeesatisfaction with particular reference to a police forceordm Total Quality Management Vol 9pp 235-47
Appendix The complete satisfaction modelThe general form of the inner structural model is
Z = bZ + gx + n E(njx) = 0 (A1)
where Z = (Z1Z2 Z7) represents the vector of endogenous latent variables x = (x1 x2) thevector of exogenous latent variables b and g are parameter matrices of suitable order and n isthe error term
EJM371112
1720
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
The equations of the model represented in Figure 1 are
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 b32 0 0 0 0 0 0
b41 0 b43 0 0 0 0 0
b51 0 b53 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 b82 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 g22
0 g32
0 g42
0 g52
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A2)
where x1 plusmn image x2 plusmn perceived employee satisfaction Z1 plusmn customer expectations Z2 plusmnperceived employee loyalty Z3 plusmn perceived employee commitment Z4 plusmn perceived productquality Z5 plusmn perceived service quality Z6 plusmn perceived value Z7 plusmn customer satisfaction Z8 plusmncustomer loyalty
It should be noted that the ECSI model can easily be derived from equation (2) if we excludethe impacts of the three new latent variables x2 Z2 and Z3 ie
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
=
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
b41 0 0 0 0 0 0 0
b51 0 0 0 0 0 0 0
b61 0 0 b64 b65 0 0 0
b71 0 0 b74 b75 b76 0 0
0 0 0 0 0 0 b87 0
2
66666666666666664
3
77777777777777775
Z1
Z2
Z3
Z4
Z5
Z6
Z7
Z8
2
66666666666666664
3
77777777777777775
+
g11 0
0 0
0 0
0 0
0 0
0 0
g71 0
g81 0
2
66666666666666664
3
77777777777777775
x1
x2
+
n1
n2
n3
n4
n5
n6
n7
n8
2
66666666666666664
3
77777777777777775
(A3)
The measurement model relating the latent variables to the manifest variables has the generalform
y = LyZ + e x = Lxx + d E(e) = E(d) = E(ejZ) = E(djx) = 0 (A4)
Employee-customer
satisfaction
1721
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722
where y = (y1 y2 yp) and x = (x1 x2 xq) are the manifest endogenous and exogenousvariables respectively Ly and Lx are the corresponding parameter matrices
Representing by yi = (yi1 yiH i
) the vector of manifest variables related to the latentendogenous variable Zi and by x
i = (xi1 xiGi) the vector of manifest variables related to the
latent exogenous variable xi we can also write the model in the form
yij = lyijZi + e ij i = 1 7 j = 1 H i xij = lxijxi + dij i = 1 2
j = 1 Gi
(A5)
where Hi is the number of manifest variables associated with variable Zi and Gi is the number ofmanifest variables associated with variable xi
EJM371112
1722